convergence problem in big magnetic system

Message

okuno · #1 Post by **okuno** » Fri May 18, 2007 2:40 am

Dear Vasp Users and administrators.

and calculate about 10 cycle and then use the wave function
and set mixing term default and restart the calculation.
But the results is oscilatte.

How we treat such big magnetic (insulator) system to
converge? We should reduce AMIX BMIX more?

My input file,
ISTART=1
ENCUT=600
PREC=High
ISMEAR=0
ISPIN=2
ICHARG=2
VOSKOWN=1
EDIFF=1e-5
MAGMOM=0 0 1 1 1 1 -1 -1 -1 -1 -1 -1 -1 -1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

#AMIX=0.2
#BMIX=0.0001
#AMIX_MAG=0.2
#BMIX_MAG=0.0001

NELM=100

LWAVE=.TRUE.
LREAL=Auto

My KPOINTS is as
Automatic mesh
0
Gamma
4 4 2
0. 0. 0.

the energy is oscilate as,
N E dE d eps ncg rms rms(c)
DAV: 1 -0.641280906003E+03 -0.64128E+03 -0.29672E+02 18128 0.103E+02 0.817E+01
DAV: 2 -0.871303715059E+03 -0.23002E+03 -0.10488E+03 17096 0.187E+02 0.176E+02
DAV: 3 -0.643445537167E+03 0.22786E+03 -0.42828E+02 14242 0.192E+02 0.111E+02
DAV: 4 -0.546437299682E+03 0.97008E+02 -0.19717E+02 16738 0.117E+02 0.711E+01
DAV: 5 -0.538404786824E+03 0.80325E+01 -0.69648E+01 16784 0.507E+01 0.640E+01
DAV: 6 -0.529293144140E+03 0.91116E+01 -0.12544E+01 16152 0.142E+01 0.379E+01
DAV: 7 -0.532029837949E+03 -0.27367E+01 -0.94502E+00 16264 0.199E+01 0.518E+01
DAV: 8 -0.527410735292E+03 0.46191E+01 -0.22945E+00 16120 0.105E+01 0.355E+01
DAV: 9 -0.527674969029E+03 -0.26423E+00 -0.37772E-01 14296 0.278E+00 0.372E+01
DAV: 10 -0.529752826762E+03 -0.20779E+01 -0.42087E-01 17976 0.453E+00 0.436E+01
DAV: 11 -0.531909605700E+03 -0.21568E+01 -0.20232E+00 18344 0.101E+01 0.440E+01
DAV: 12 -0.531325259746E+03 0.58435E+00 -0.11588E+00 17752 0.560E+00 0.430E+01
DAV: 13 -0.530750161427E+03 0.57510E+00 -0.42769E-01 17880 0.281E+00 0.408E+01
DAV: 14 -0.530140733334E+03 0.60943E+00 -0.52814E-01 18816 0.462E+00 0.342E+01
DAV: 15 -0.530033402657E+03 0.10733E+00 -0.72433E-01 18608 0.513E+00 0.279E+01
DAV: 16 -0.530166483415E+03 -0.13308E+00 -0.28359E-01 17192 0.262E+00 0.271E+01
DAV: 17 -0.530316475947E+03 -0.14999E+00 -0.90737E-02 18024 0.205E+00 0.273E+01
DAV: 18 -0.530539280787E+03 -0.22280E+00 -0.42134E-01 18808 0.466E+00 0.238E+01
DAV: 19 -0.530645785554E+03 -0.10650E+00 -0.59267E-01 18584 0.465E+00 0.213E+01
DAV: 20 -0.530125894996E+03 0.51989E+00 -0.98119E-01 18952 0.767E+00 0.171E+01
DAV: 21 -0.529519810189E+03 0.60608E+00 -0.63536E-01 17112 0.445E+00 0.128E+01
DAV: 22 -0.529310464899E+03 0.20935E+00 -0.36594E-01 18536 0.507E+00 0.115E+01
DAV: 23 -0.529434483273E+03 -0.12402E+00 -0.42629E-01 18184 0.301E+00 0.115E+01
DAV: 24 -0.529522910968E+03 -0.88428E-01 -0.15513E-01 17032 0.297E+00 0.899E+00
DAV: 25 -0.529443173809E+03 0.79737E-01 -0.15178E-01 18512 0.297E+00 0.612E+00

<span class='smallblacktext'>[ Edited ]</span>

#2 Post by **admin** » Fri May 18, 2007 1:23 pm

the information you give is a liitle too unspecific to give detailled hints.
1) large magnetic systems are hard to converge sometimes: The oscillations are damped, so your run may converge finally. It may be necessary to increase NELM for the first ionic step(s) to reach the desired EDIFF.
2) please also choose the MAGMOM input as good as possible
3) it may also help to use different PPs (eg the PAW-PBE ones, unless you have used them for this run already)

okuno · #3 Post by **okuno** » Mon May 21, 2007 2:12 am

Dear asministrator, thank you for your advice.

I run the SCF calculation 100 NELM and use PAW-PBE, but the result did not converged like,

DAV 92: -0.524276430207E+03 -0.94874E-01 -0.17360E-01 4072 0.133E+00 0.198E+01
DAV: 93 -0.524108972227E+03 0.16746E+00 -0.56939E-02 4320 0.156E+00 0.198E+01
DAV: 94 -0.524107599498E+03 0.13727E-02 -0.14779E-02 4168 0.684E-01 0.193E+01
DAV: 95 -0.524099376685E+03 0.82228E-02 -0.10181E-02 4120 0.498E-01 0.192E+01
DAV: 96 -0.524954677557E+03 -0.85530E+00 -0.58180E-01 4592 0.563E+00 0.159E+01
DAV: 97 -0.525434711165E+03 -0.48003E+00 -0.61430E-01 4552 0.518E+00 0.154E+01
DAV: 98 -0.525794195355E+03 -0.35948E+00 -0.27443E-01 4448 0.315E+00 0.193E+01
DAV: 99 -0.526570008614E+03 -0.77581E+00 -0.19807E-01 4222 0.304E+00 0.206E+01
DAV: 100 -0.526268528579E+03 0.30148E+00 -0.68352E-02 4318 0.168E+00

Are there any room to converge the such a big magnetic calculation ,
for example reduce AMIX or AMIXMAG ?

okuno · #4 Post by **okuno** » Mon May 21, 2007 4:37 am

In addition, our treating system is magnetic ion is Fe and ferrite
system. The crystal symmetry is P63/mmc SG number 194 hexagonal . There are 24 Fe ion with 8 down spin and 16 up spin

zhichengzhong · #5 Post by **zhichengzhong** » Mon Jun 04, 2007 12:01 pm

I also meet the similar problem.
My suggestion is :
1,How about ISMEAR=-5 ?
2, does non-magnetic calculation converge(ISPIN=1)? If does, then you use the CHGCAR and WAVECAR to do the mag calculation.
Good Luck